The impact of air transport delays and their propagation has long been studied, mainly from environmental and mobility viewpoints, using a wide range of data analysis tools and simulations. Less attention has nevertheless been devoted to how delays create meso-scale structures around each airport. In this work we tackle this issue by reconstructing functional networks of delay propagation centred at each airport, and studying their identifiability (i.e. how unique they are) using Deep Learning models. We find that such delay propagation neighbourhoods are highly unique when they correspond to airports with a high share of Low Cost Carriers operations; and demonstrate the robustness of these findings for the EU and US systems, and to different methodological choices. We further discuss some operational implications of this uniqueness.